You may try for ANFIS with Subtractive Clustering, over ANN for better RMSE Values, which is one of the measure to judge the accuracy of model.
ANFISEDIT is the command to be used to evoke GUI of ANFIS.
You may try clustering with different cluster radii.
Regards
Ashu S. Kedia
Ur question is a complete course on ANN. I will recommend going through mathematical equations used in training an ANN as improving RMSE depends on many factors e.g no of training examples, their accuracy, pre/post processing, traing algo , no of neurns, hidden layers to name few. Also, the performance parameters are inbuilt in all ANN models/ architectures in MATlab as structures for which u need to see the content file and help commands.
The final error value learning is stored in the parameter net.trainParam.goal. You can change it by substitution, eg .: net.trainParam.goal = 0.001
Regards
In addition to Imran Shafi and Jacek Bartman suggested parameters you have to also consider the values of mu and learning rate. More importantly training algorithm itself also has influence on error-different algorithm work differently with a particular problem.
- Badges
- Science topic